I'll ask on AWS. Can you just kind of go through how you're feeling about your capacity levels and how capacity constrained you are right now? And then in your prepared remarks, you mentioned Trainium3 demand and maybe broadening out your customer base. Can you talk about the demand you're seeing outside of your major customers for Trainium? Andrew Jassy (CEO): Yes. We have significantly increased our capacity, bringing in 3.8 gigawatts over the past year, with an additional gigawatt expected in the fourth quarter. We anticipate doubling our total capacity by the end of 2027. Currently, the industry faces potential bottlenecks with power, and possibly chips in the future, but we are effectively monetizing the capacity we have. Regarding Trainium demand, it's performing well, particularly Trainium2, which is fully subscribed and has become a multibillion-dollar business, experiencing a 150% increase in revenue quarter-over-quarter. Large projects like Project Rainier with Anthropic are utilizing Trainium2 to train the next version of their AI model, scaling from 500,000 to 1 million chips by the end of the year. While we have a limited number of substantial customers for Trainium2, its price performance, which is 30% to 40% better than alternatives, makes it attractive for customers focusing on AI and inference workloads. There is strong demand for Trainium, and we expect to preview Trainium3 at the end of this year, with more supply in early 2026. Many large and medium-sized customers have shown significant interest in Trainium3. null (Operator): And the next question comes from the line of Brian Nowak with Morgan Stanley.

Congrats on the quarter, guys. So maybe 2. One, Andy, sort of a philosophical chip question. There's a lot of questions in the market about Trainium and sort of its positioning versus other third-party chips. So how do you think about the key hurdles of Trainium3 need to overcome to really make Trainium adoption broader to your point on the last question and continue to drive Trainium, as opposed to satisfying what could be broader demand with third-party chips in the near term? Andrew Jassy (CEO): Yes. Well, first of all, we're always going to have multiple chip options for our customers. It's been true in every major technology building block or component that we've had in AWS. Really in the history of AWS, it's never just one player that over a long period has the entire market segment and then can satisfy everybody's needs on every dimension. And so we have a very deep relationship with NVIDIA. We have for a very long time, and we will for as long as I can foresee the future. We buy a lot of NVIDIA. We are not constrained in any way in buying NVIDIA, and I expect that we'll continue to buy more NVIDIA both next year and in the future. But we're different from most technology companies in that we have our own very strong chip team, and this is our Annapurna team. And you saw it first on the CPU side with what we built with Graviton which is about 40% better price performance than the other x86 processors, and you're seeing it again on the custom silicon on the AI side with Trainium, which is about the same amount of price performance benefit for customers relative to other GPU options. And our customers to be able to use AI as expansively as they want. And remember, it's still relatively early days at this point. They're going to need better price performance and they care about it deeply. And so I mentioned earlier the momentum that Trainium2 has. And I think that for us, as we think about Trainium3, I expect Trainium3 will be about 40% better than Trainium2 and Trainium2 is already very advantaged on price performance. So we have to, of course, deliver the chip. We have to deliver it in volumes and deliver it quickly, and we have to continue to work on the software ecosystem, which gets better all the time. And as we have more proof points like we have with Project Rainier with what Anthropic is doing on Trainium2, it builds increasing credibility for Trainium. And I think customers are very bullish about it. I'm bullish about it as well. null (Operator): And our next question comes from the line of Doug Anmuth with JPMorgan.

I'll stick with basically the same topic, Andy. But can you just talk a little bit about the architecture of Project Rainier and how it's differentiated and what that means for customers and for AWS? And do you expect Rainier to expand beyond Anthropic? And how do you replicate Rainier with Trainium3 chips? Andrew Jassy (CEO): Yes. I think what's compelling for Anthropic regarding Project Rainier is really the Trainium2 chip. We've built a very large cluster that they can use extensively. Constructing a cluster with over 500,000 chips, potentially reaching 1 million, is a significant infrastructure achievement that is challenging to replicate at scale. Part of this is due to the infrastructure capabilities we've developed over an extended period in AWS, which is unique in the industry. Additionally, the chip's performance and cost-effectiveness are important factors. While Project Rainier is tailored for Anthropic, we have many other customers interested in utilizing large clusters of Trainium chips, and we hope to provide them the opportunity with Trainium3. null (Operator): The next question comes from the line of Mark Mahaney with Evercore ISI.

groceries and future headcount considerations. Regarding groceries, last quarter you mentioned that around 70% of users had never purchased perishables from Amazon before. Can you elaborate on whether this presents a significant shift for you? You referred to it as a potential game changer earlier. Does this imply that we may not need Amazon Fresh stores anymore? You've always had an advantage with your delivery network. Have you reached a point where you can truly shift consumer habits and position Amazon as a primary choice for groceries? Do you believe you are at that stage now? On the second topic of headcount, given the recent news, could you share your perspective on how you see headcount evolving? Are the efficiencies gained from AI substantial enough to maintain a stable headcount for the near future? Please discuss the advantages and disadvantages in terms of headcount moving forward. Andrew Jassy (CEO): Yes, I'll begin with grocery. We have a substantial grocery business. Excluding Whole Foods Market and Fresh, our grocery business generated over $100 billion in gross merchandising sales in the last year, placing us among the top three grocers in the U.S. A significant portion of this includes items typically found in the middle aisle, such as consumables, canned goods, pet food, and health and beauty products, all of which are growing rapidly. Additionally, Whole Foods Market, a leader in organic foods, is also experiencing growth at a pace exceeding most grocery companies, along with promising profitability. We plan to increase our Whole Foods locations in the coming years. I am also excited about our new concept, Daily Shop, which offers a smaller version of Whole Foods in urban areas. We have launched three of these locations, which have begun successfully, and you can expect to see more. We've consistently aimed to enhance our physical presence, and we continue to explore various formats. The initiative we are most enthusiastic about is our capability to provide perishable groceries with same-day delivery. Many of our customers shop multiple times a week, purchasing items like shampoo, detergent, or water. The ability for them to add perishables like milk, eggs, and yogurt to their order, all from the same cart, and have it delivered within a few hours is very appealing. About a year ago, we initiated this in a few markets and were pleasantly surprised by the quick uptake, not just in the number of customers buying perishables, but also in how frequently they returned for future grocery purchases. We have now expanded this service to 1,000 cities across the U.S., with plans to reach 2,300 by the end of this year, which is significantly altering the trajectory and scale of our grocery business. I believe the traditional model of weekly grocery stock-ups is evolving, and we are playing a key role in that change. While we will continue to experiment with other physical formats, our advancements in perishables, especially through our same-day delivery capabilities, are highly significant. Regarding headcount, the recent announcement we made was not primarily about financial or AI concerns. It is more about our culture. Rapid growth over the years has led to increased staff, locations, and business types, resulting in more layers within the organization. This can unintentionally diminish the ownership among our employees who handle the day-to-day operations and make quick decisions. As a leadership team, we are dedicated to functioning like the world's largest startup, which involves streamlining operations, enhancing employee ownership, and fostering innovation and agility. With the current technological transformation, being lean and responsive is crucial, and that is our focus moving forward. null (Operator): And the next question comes from the line of Eric Sheridan with Goldman Sachs.

Wanted to know, Andy, if you could reflect on the opportunity that's continuing to present itself in terms of rolling out more robotics and automation and the broader theme of physical AI across your operations? And how should we be thinking about that as a driver of potential efficiencies, but also as a driver of the ability to possibly reinvest back in the business over the long term? Andrew Jassy (CEO): Robotics is a very substantial area of investment for us. We have over 1 million robots in our fulfillment network at this point. I would say that while that's significant, we have a lot of invention in flight. So I expect that we'll have more over a period of time. Robotics are very important for us and for our customers and for our teammates because they improve safety, they boost productivity, they increase speed, and they let our human teammates focus on problem solving and what they do best. We expect that our people remain at the heart and center of our fulfillment network as they have from when we first started working with robotics. We expect that over time, we will have a fulfillment network where robots and humans complement each other and work together. You're going to continue to see us advance and invest very significantly in robotics. It's going to help on the safety, the productivity, the speed, and ultimately some of the cost pieces, which will allow us to continue to improve the customer experience. null (Operator): And the next question comes from the line of John Blackledge with TD Cowen.

How does Amazon think about agentic commerce going forward? And how do you think Amazon will serve customers using agents to purchase goods on Amazon in the future? Andrew Jassy (CEO): I'm very excited about the long-term potential of agentic commerce. It has the potential to benefit customers and significantly enhance e-commerce. If you know what you want to buy, there's not much better than shopping on Amazon. However, if you're unsure about your purchase, a physical store and a salesperson still offer some advantages. Many people use Amazon successfully, but it's often helpful to ask questions for guidance on what to look for. As you ask more questions and have various options presented, I believe AI and agentic commerce will improve the online experience, making it easier to narrow down choices compared to traditional retail settings. We are actively working on our own agentic commerce initiatives, including Rufus, which is improving and expanding in usage. We also offer features like Buy for Me, where we can find items on Amazon that we don't carry but can purchase from other merchants for customers. Both initiatives have been successful. Over time, we also plan to collaborate with third-party agents. This reminds me of the early days of search engines, which became invaluable for commerce discovery. Currently, search engines contribute only a small portion of our referral traffic, and third-party agents represent an even smaller fraction, but we are committed to forming partnerships that enhance the customer experience. Right now, the customer experience lacks personalization, shopping history, accurate delivery estimates, and correct pricing. We need to improve these aspects and ensure proper value exchange. The exciting part is that AI and agentic commerce solutions will increase online shopping, which is beneficial for customers and for Amazon as it will allow us to offer the widest selection, great value, and reliable fast delivery. This bodes well for our future. null (Operator): And our final question comes from the line of Colin Sebastian with Baird.

I guess first on AWS, following up there. How much of this acceleration is driven by core infrastructure versus AI workload monetization? And I think part of it is trying to understand how important newer services like AgentCore are becoming and bringing enterprises to AWS to build agents? And then I guess, secondly, regarding the acceleration in advertising, if you could potentially disaggregate the core advertising contribution versus DSP and Prime video. That would be helpful as well. Andrew Jassy (CEO): I'll start with AWS. We are very pleased with this quarter's results, showing a 20% year-over-year growth on an annualized run rate of $132 billion, which is impressive. There is clear momentum. We are experiencing growth in our AI segment, particularly in inference, training, and the use of our Trainium custom silicon. Bedrock and SageMaker are expanding rapidly, and many companies are now focusing on building agents. I believe that much of the future value in AI for these companies will come from agents. However, developing agents remains challenging, which is why we created strands, an open-source tool that allows users to build agents from any model they envision. Customers who prioritize security and scalability are beginning to adopt agents but have lacked the foundational components necessary for secure and scalable agents that can support their businesses and enhance customer experiences and data management. This is why we developed AgentCore, which offers essential building blocks similar to what we provided in the early days of AWS with compute, storage, and databases. AgentCore offers the components needed to deploy agents securely and efficiently, and we have found that this resonates strongly with our customers. There is nothing else like it, and it is altering their timelines and willingness to create agents, which is very compelling for them. We are also seeing growth in AWS from enterprises transitioning from on-premises infrastructure to the cloud, and we continue to capture a significant share of those transformations. Given our current momentum, I believe we can maintain this growth rate for some time. On the advertising front, we are also very satisfied with our progress. Every one of our advertising services showed meaningful growth this quarter. Several factors contribute to this. We offer a comprehensive advertising solution that spans from widespread awareness initiatives at the top of the funnel to targeted strategies at the point of sale with sponsored products. Few companies can match the robustness of our full funnel approach. Additionally, our capabilities in audience curation and measurement provide a return on advertising spend that is quite distinct. There are several areas where we see ongoing growth. In our stores business, a large portion of retail—80% to 85%—still occurs in physical locations. Over time, that will change, and I believe AI will accelerate this transition. We have substantial opportunities in our existing stores and with video advertising, which has already yielded significant revenue, although we are still in the early stages. This segment is likely to continue being a major growth area. Our demand-side platform, Amazon DSP, is also growing rapidly. We have addressed customer feedback over the past 20 months and closed key gaps, making our DSP fully featured. Our partnership with Roku provides the largest connected TV presence in the U.S. Furthermore, we have added integration opportunities with ad inventory from Netflix, Spotify, and SiriusXM for our DSP customers. This combination is powerful, and our demand-side platform is expanding quickly, giving us a positive outlook on our progress while noting that we still have work to do. However, we are far from reaching our growth potential there. null (Operator): Thanks for joining us on the call today and for your questions. A replay will be available on our Investor Relations website for at least 3 months. We appreciate your interest in Amazon and look forward to speaking with you again next quarter.